Thermal analysis of Cretaceous ambers from southern France

Eugenio RAGAZZI University of Padua, Department of Pharmacology and Anaesthesiology, largo E. Meneghetti 2, I-35131 Padova (Italy) [email protected]

Aurelio GIARETTA Institute of Geosciences and Earth Resources (IGG-CNR), corso Garibaldi 37, I-35137 Padova (Italy)

Vincent PERRICHOT Paleontological Institute, University of Kansas, Lindley Hall, 1475 Jayhawk blvd, Lawrence, KS 66045 (USA) [email protected]

Didier NÉRAUDEAU Université Rennes I, UMR CNRS 6118 Géosciences, campus de Beaulieu bât. 15, 263 avenue du Général Leclerc, F-35042 Rennes cedex (France) [email protected]

Alexander R. SCHMIDT Georg-August-Universität Göttingen, Courant Research Centre Geobiology, Goldschmidtstr. 3, D-37077 Göttingen (Germany) and Museum für Naturkunde der Humboldt-Universität zu Berlin, Invalidenstr. 43, D-10115 Berlin (Germany) [email protected]

Guido ROGHI Institute of Geosciences and Earth Resources (IGG-CNR), via Matteotti 30, I-35137 Padova (Italy) [email protected]

Ragazzi E., Giaretta A., Perrichot V., Néraudeau D., Schmidt A. R. & Roghi G. 2009. — Thermal analysis of Cretaceous ambers from southern France. Geodiversitas 31 (1) : 163-175.

ABSTRACT Th ermal properties of French Cretaceous ambers were investigated and compared with other ambers from various sites of the world. Th e amber samples came

GEODIVERSITAS • 2009 • 31 (1) © Publications Scientifi ques du Muséum national d’Histoire naturelle, Paris. www.geodiversitas.com 163 Ragazzi E. et al.

from 10 diff erent localities in southern France, in the Charentes, Languedoc, and Provence regions, ranging from Late Albian to Santonian in age. Th ermo- gravimetric (TG) and Diff erential Th ermogravimetric (DTG) profi les were obtained at heating rate of 10 K/min in air, starting from room temperature (20°C) and reaching a maximum temperature of 700°C. Elemental Analysis for total Carbon, Hydrogen, Nitrogen and Sulphur was also carried out. Th e TG combustion profi le of the resins started after 200°C and complete combustion took place near 600°C. Th e DTG behaviour is characterized by a main exo- thermal peak situated between 394 and 420°C, accompanied by minor peaks and shoulders. Th e increasing value of the main exothermal peak correlates well to the increase of the age of the specimens, with a signifi cant correlation coeffi cient (r = 0.7721, p = 0.0089). A signifi cant correlation (r = 0.6728, p = 0.0004) is also found with other samples of diff erent age and origin. By con- sidering the whole pattern of DTG peaks, a possible fi ngerprinting model of the French ambers is evaluated by multivariate analysis. Cluster Analysis and Principal Component Analysis show the presence of several clusters, according to the geological age and possibly to the palaeobotanical origin. Th e elemental analysis is consistent with that of other Cretaceous samples from diff erent sites of the world. Carbon and hydrogen are the main constituents (range 73-80% and 9.5-11.5% respectively). Sulphur is detected in small amounts (0.8-2.4%). KEY WORDS Nitrogen is absent or appears as traces only (0-0.008%). Oxygen and other ele- Th ermal analysis, ments range from 4.6 to 16.8%. No successful clustering was possible according elemental analysis, to the elemental composition. Th ermal analysis, completed with multivariate amber, Cretaceous, statistics, is a useful source of information also for French ambers, as a help for France. identifi cation of the age, diagenetic processes and palaeobotanical origin.

RÉSUMÉ Analyse thermique des ambres crétacés du Sud de la France. Les propriétés thermiques des ambres crétacés de France sont évaluées et comparées avec celles d’autres ambres de sites variés. Les échantillons d’ambre proviennent de 10 localités diff érentes du Sud de la France, dans la région des Charentes, en Languedoc, et en Provence, et ont un âge compris entre l’Albien terminal et le Santonien. Les profi ls thermo-gravimétriques (TG) et thermo-gravimétriques diff érentiels (DTG) ont été obtenus à un taux de réchauff ement de 10 K/min dans l’air, débutant à température ambiante (20°C) et atteignant une tempéra- ture maximale de 700°C. L’analyse élémentaire du total de carbone, hydrogène, azote et soufre a également été eff ectuée. Le profi l TG des résines débute après 200°C et la combustion complète se situe autour de 600°C. Le comportement DTG est caractérisé par un pic exothermique principal situé entre 394 et 420°C, ainsi que par des pics et variations mineurs. Les valeurs croissantes du principal pic exothermique et de l’âge des échantillons d’ambre montrent un coeffi cient de corrélation signifi catif (r = 0,7721, p = 0,0089). Une bonne corrélation (r = 0,6728, p = 0,0004) est aussi observée avec d’autres ambres d’origine et d’âge distincts. Une analyse statistique multivariée considérant l’ensemble des pics DTG permet d’établir un profi l caractéristique des ambres français. Une analyse de clusters et une analyse en composantes principales montrent plu- sieurs regroupements en accord avec l’âge géologique et possible ment l’origine paléo botanique des résines. La composition élémentaire est cohérente avec celle d’autres ambres crétacés. Carbone (73-80 %) et hydrogène (9,5-11,5 %) sont les principaux constituants, alors que le soufre est présent en petite quantité (0,8-2,4 %) et que l’azote est absent ou présent seulement à l’état de traces

164 GEODIVERSITAS • 2009 • 31 (1) Th ermal analysis of Cretaceous French amber

(0-0,008 %). L’oxygène et les éléments traces varient entre 4,6 et 16,8 %. La MOTS CLÉS composition élémentaire ne permet pas d’établir de regroupements signifi catifs Analyse thermique, des diff érents ambres analysés. L’analyse thermique, complétée par des analyses analyse élémentaire, statistiques multivariées, s’avère donc informative tant du point de vue de l’âge ambre, Crétacé, que de celui des processus diagénétiques et de l’origine paléobotanique, et per- France. met d’affi ner la caractérisation des ambres de France.

INTRODUCTION from diff erent geographical sources (Broughton 1974; Rodgers & Currie 1999; Ragazzi et al. Amber is a fossil organic polymer which origi- 2003; Feist et al. 2007). It is a rapid quantitative nates from tree resins through complex matura- method which aims to examine the overall pyrolysis tion processes collectively named “amberization” process and then to estimate the eff ective rates of (Anderson & Winans 1991; Anderson et al. 1992; overall decomposition reactions. Th e fi rst studies Anderson & Crelling 1995). Th is process includes (Broughton 1974; Rodgers & Currie 1999) have oxidation, oligomerization and cross-linking, which demonstrated the ability of thermal analysis to take place in the resin soon after its secretion by detect specifi c characteristics of various kinds of the plant. Several diff erent parameters can play recent and fossil resins, with the oldest samples relevant roles in amber composition, for example from the Eocene (about 40-45 Ma) and the Late the original environment (as a complex mixing of Cretaceous (about 70 Ma). Further investiga- temperature, light irradiation, climate and possible tions conducted in our laboratory extended the pathogen intervention, which infl uence the resin application of the thermal analysis to more an- composition), the palaeobotanical origin, the dia- cient fossil resins, till the Upper Triassic (about genetic history, the geological age, and also possible 225 Ma; Ragazzi et al. 2003). Under diff erential alterations which occurred after amber mining. thermogravimetric (DTG) analysis, all samples Th e characterization of amber can be achieved presented a main exothermal event, whose tem- using several methods of analysis, such as Fourier perature varied among resins, showing signifi cant transformed infra-red analysis (FTIR), pyrolysis-gas linear correlation, on the basis of the geological chromatography-mass spectrometry and nuclear age (Ragazzi et al. 2003). magnetic resonance (Broughton 1974; Lambert & Th e availability of recently discovered Albian Frye 1982; Beck 1986; Grimalt et al. 1988; An- to Santonian (about 100 to 85 Ma) amber from derson & Winans 1991; Anderson et al. 1992; France (Néraudeau et al. 2002, 2003, 2008, 2009 Anderson & Crelling 1995; Carlsen et al. 1997; this volume; Gomez et al. 2003; Perrichot 2005; Martínez-Richa et al. 2000). Th e chemical analysis Perrichot et al. 2007a, b), suggested an investiga- of the basic components of the polymer may fail tion of the thermal properties of these peculiar to reveal diff erences among fossil resins since the fossil resins, and a comparison with the behaviour specimens, although of the same geological age of other ambers from various sites of the world. and similar composition, may display diff erent In particular, the mid-Cretaceous ambers from maturity grades, for example as a consequence of Charentes region, whose deposits are the largest higher temperatures and chemical linkages which in France, were considered. Besides thermogravi- occurred during fossilization. metric investigation, an elemental analysis was Th ermogravimetric analysis has been introduced also carried out on the same samples in order as a new method to provide additional informa- to further defi ne the amber physico-chemical tion on the physico-chemical properties of amber characteristics.

GEODIVERSITAS • 2009 • 31 (1) 165 Ragazzi E. et al.

TABLE 1. — Characteristics of the French amber samples investigated. Abbreviations in brackets are those used elsewhere in this paper.

Sample Origin Colour Geological age Main DTG Other DTG no. and thermal peak thermal peaks label (°C) (°C) 1 (Arc) Archingeay Opaque dark yellow Late Albian 408 453, 574, (Charente-Maritime) (Dispar zone, 600 100.6-99.6 Ma) 2 (Cdl) Cadeuil Opaque dark yellow Late Albian 408 473, 586 (Charente-Maritime) (Dispar zone, 100.6-99.6 Ma) 3 (Frs) Fouras Light yellow Early Cenomanian 420 347, 557, (Charente-Maritime) (Mantelli zone, 578, 600 99.6-98.6 Ma) 4 (Aix) Île d’Aix Light yellow Early Cenomanian 420 341, 538, (Charente-Maritime) (Mantelli zone, 578, 616 99.6-98.6 Ma) 5 (Buz) La Buzinie Opaque dark Early Cenomanian 411 328, 427, (Charente) yellow/orange (Mantelli zone, 584 99.6-98.6 Ma) 6 (Fou) Fourtou Dark yellow/orange Cenomanian 415 321, 544 (Aude, Eastern Pyrenees) (99.6-93.5 Ma) 7 (Sal) Salignac Dark yellow/orange Cenomanian 411 334, 427, (Alpes de Haute-Provence) (99.6-93.5 Ma) 578 8 (Plc) Piolenc Light yellow Santonian 397 334, 380, (Vaucluse) (85.8-83.5 Ma) 578, 600 9 (Bel) Belcodène Light yellow Santonian 406 557 (Bouches-du-Rhône) (85.8-83.5 Ma) 10 (Ens) Ensuès-la-Redonne Dark yellow/orange Santonian 394 446, 557 (Bouches-du-Rhône) (85.8-83.5 Ma)

MATERIALS AND METHODS available at http://www.stratigraphy.org/upload/ ISChart2008.pdf) and Gradstein et al. (2004). Th e amber samples analysed herein come from 10 diff erent localities in Southern France and their THERMOGRAVIMETRIC (TG) AND DIFFERENTIAL main characteristics are given in Table 1. Samples THERMOGRAVIMETRIC ANALYSIS (DTG) no. 1-5, coming from Charente-Maritime and Th ermogravimetric (TG) and Diff erential Th er- Charente, originate from diff erent deposits of mogravimetric (DTG) profi les were obtained for Late Albian and Early Cenomanian age and are each sample using a prototypal C.N.R. instrument of diff erent varieties of colour and transparency. (IGG-CNR, Padua, Italy) named “Le Chatelier”. A Th eir geological age was attributed according type S (Pt-10% Rh/Pt) thermocouple placed inside to palynological (dinofl agellate cysts), micro- an electric furnace, provided sample and furnace paleontological (foraminifers and ostracods), or measurements. When possible, samples were taken stratigraphical data as provided in former publica- from the internal part of the pieces of resin, in order tions (Néraudeau et al. 2002, 2003, 2008, 2009 to avoid the infl uence of degradation processes oc- this volume; Gomez et al. 2003; Perrichot et curred in the surface, or to limit the infl uence of a al. 2004, 2007a; Dejax & Masure 2005; Peyrot concentration gradient of the diff erent chemical spe- et al. 2005). Th e absolute age was considered ac- cies present in the sample. Samples were pulverized cording to the International Stratigraphic Chart in an agate mortar before the measurement (mass (International Commission on Stratigraphy 2007, 500 mg, particle size < 75 μm) and were inserted in

166 GEODIVERSITAS • 2009 • 31 (1) Th ermal analysis of Cretaceous French amber

TABLE 2. — Elemental analysis of the resin samples. Each value is the mean ± sd of duplicate determinations. Abbreviations refer to samples as indicated in Table 1.

Sample no. and label Carbon (%) Hydrogen (%) Sulphur (%) Nitrogen (%) Oxygen and other elements (%) 1 (Arc) 78.65 ± 0.27 10.92 ± 0.08 0.38 ± 0.08 0.005 ± 0.007 10.05 ± 0.26 2 (Cdl) 77.43 ± 0.34 10.80 ± 0.02 0.33 ± 0.09 0 11.44 ± 0.41 3 (Frs) 79.69 ± 0.60 10.91 ± 0.01 0.46 ± 0.01 0 8.95 ± 0.60 4 (Aix) 80.15 ± 0.06 10.93 ± 0.07 0.59 ± 0.01 0 8.33 ± 0.13 5 (Buz) 77.27 ± 0.45 10.75 ± 0.01 1.12 ± 0.06 0 10.86 ± 0.50 6 (Fou) 77.79 ± 0.65 10.47 ± 0.10 2.40 ± 0.02 0 9.33 ± 0.73 7 (Sal) 73.32 ± 1.76 9.55 ± 0.32 0.37 ± 0.01 0.008 ± 0.012 16.76 ± 2.08 8 (Plc) 83.14 ± 4.88 11.47 ± 0.75 0.81 ± 0.08 0 4.59 ± 5.71 9 (Bel) 78.23 ± 0.30 10.48 ± 0.04 0.98 ± 0.01 0.002 ± 0.003 10.31 ± 0.25 10 (Ens) 78.21 ± 0.70 10.50 ± 0.06 1.58 ± 0.01 0.004 ± 0.005 9.71 ± 0.74 Mean values ± sd 78.39 ± 2.71 10.68 ± 0.52 0.90 ± 0.65 0.002 ± 0.005 10.03 ± 3.27 a platinum crucible, placed on quartz glass support Th e powdered sample (about 2 mg) was placed interfaced by Mettler Toledo AB 104 balance. Th e into a tin capsule. Th e sample was loaded into the heating rate was 10 K/min in air, starting from room AS 200 autosampler which allows the automatic temperature (20°C) and reaching a maximum tem- analysis. Th e calibration standards for carbon- perature of 700°C. Th e ash residue of the samples hydrogen-nitrogen and sulphur were prepared from was 3.45% ± 1.66 (mean ± sd), and the range was known amounts of sulfanilamide (C6H8N2O2S). 0.74% (sample no. 3) to 5.4% (sample no. 6). A Analytical results for total carbon, hydrogen, nitro- computer equipped by custom-made software, writ- gen and sulphur in the amber samples obtained from ten in Lab View 5.1 language, recorded sequential at least two determinations are shown in Table 2. temperature and weight of sample (TG signal) and derived DTG data from the recorded TG signal. DATA PROCESSING AND STATISTICAL ANALYSIS Th e TG and DTG profi les were edited using the To examine whether a link between the main DTG Graphing Software GRAPHER version 2. thermal peak and the geological age was present, a scatter-plot was constructed and the Pearson’s ELEMENTAL ANALYSIS correlation coeffi cient (r) was calculated. Th e level A CE-Instruments EA 1110 Automatic Elemen- of signifi cance of the correlation coeffi cient was tal Analyser equipped with AS 200 autosampler determined using ANOVA; a value of p < 0.05 was and Mettler Toledo AT21 Comparator was used considered as signifi cant. in this study. Th e instrument is a simultaneous A linear regression line was calculated with the carbon-hydrogen-nitrogen and sulphur analyser method of least-squares; the main DTG thermal based on the reliable dynamic fl ash combustion peak was considered as the explanatory variable (x) and GC separation (with helium as carrier gas) fol- and the geological age as the dependent variable lowed by thermal conductivity detectors (TCD). (y). To allow a comparison with the characteristics A workstation permits complete automation from of other amber specimens with diff erent age and weight entry to storage of results. It consists of the geographical origin, a regression was calculated Eager 200 Software installed on compatible XT/AT using the thermogravimetric data previously de- microcomputer with a graphics printer. Th e Eager termined in our laboratory (Ragazzi et al. 2003). 200 Software off ers the possibility to use K factors While correlation calculations are symmetrical with and linear regression for calculation. Weight entry respect to x and y, and after swapping the variables via the RS 232CC link is fully automated. the same correlation coeffi cient is obtained, on

GEODIVERSITAS • 2009 • 31 (1) 167 Ragazzi E. et al.

1 (Arc) 2 (Cdl)

0 0 0 0

-10 -10 473° C -2 453° C -2 -20 600° C -20 586° C 574° C -4 -30 -30 -4 -6 -40 -40

-50 -6 -50 -8

-60 -60 -10 -8 -70 -70 -12 -80 -80 -10 408° C -14 -90 -90 408° C Weight loss (%) loss Weight -100 -12 -100 -16 Derivative (%/min) 0 100 200 300 400 500 600 700 0 100 200 300 400 500 600 700

3 (Frs) 4 (Aix) 0 0 0 0

-10 -10 -2 | 600° C -2 538° C 616° C 557° C | -20 347° C -20 341° C 578° C 578° C -30 -4 -30 -4 -40 -40

-50 -6 -50 -6

-60 -60 -8 -8 -70 -70 420° C -80 -80 -10 -10 -90 -90 420 ° C Weight loss (%) loss Weight -100 -12 -100 -12 Derivative (%/min) 0 100 200 300 400 500 600 700 0 100 200 300 400 500 600 700

5 (Buz) 6 (Fou) 0 0 0 0

-10 -10

-20 -20 -2 -2 321° C 328° C -30 584 ° C -30 544° C -40 -4 -40 -4

-50 -50

-60 -6 -60 -6

-70 -70

-80 427 ° C -8 -80 -8

-90 411° C -90 415° C Weight loss(%) -100 -10 -100 -10 Derivative (%/min) 0 100 200 300 400 500 600 700 0 100 200 300 400 500 600 700 Temperature (°C)

168 GEODIVERSITAS • 2009 • 31 (1) Th ermal analysis of Cretaceous French amber

7 (Sal) 8 (Plc) 0 0 0 0

-10 -10

-20 -2 -2 334°C -20 600°C 334 °C 578°C -30 -30 -4 -40 578°C -4 -40

-50 -50 -6 380°C -60 -6 -60

-70 -70 -8

-80 427°C -8 -80

-90 411°C -90 -10 397°C Weight loss (%) loss Weight -100 -10 -100 Derivative (%/min) 0 100 200 300 400 500 600 700 0 100 200 300 400 500 600 700

9 (Bel) 10 (Ens) 0 0 0 0

-10 -10 -2 -2 -20 -20 557°C -30 557°C -30 -4 446°C -4 -40 -40

-50 -6 -50 -6

-60 -60 -8 -8 -70 -70

-80 -80 -10 -10 -90 -90 394°C 406°C Weight loss (%) loss Weight -100 -12 -100 -12 Derivative (%/min) 0 100 200 300 400 500 600 700 0 100 200 300 400 500 600 700 Temperature (°C)

FIG. 1. — Thermogravimetric (TG, thick lines) and Differential Thermogravimetric (DTG, thin lines) analysis of French Cretaceous am- bers. Abbreviations refer to samples as indicated in Table 1. the contrary the regression calculations are not of grouping (unsupervised clustering) was used. symmetrical and the swapping with respect to x Th e DTG peak values obtained for each sample and y leads to diff erent regression lines. Th erefore, at various temperatures were inserted in a spread- both regressions were calculated and the theoreti- sheet matrix and analyzed using the software Kyplot cal regression lines presented in the scatter-plot, (version 2.0 beta 13, Yoshioka 2002) to obtain the together with the average result obtained from the hierarchical agglomerative clustering of the data. two fi tting procedures. Ward’s method of clustering was the linkage rule; the distance measure was determined using non- MULTIVARIATE ANALYSIS standardized Euclidean distance. Cluster analysis (CA) In order to fi nd useful criteria to classify the samples Principal Component Analysis (PCA) according to the whole DTG pattern, a procedure of DTG peak data were analyzed with PCA in order to hierarchical clustering without any prior knowledge detect the presence of clusters within experimental

GEODIVERSITAS • 2009 • 31 (1) 169 Ragazzi E. et al.

120 loss occurred. In all cases the TG curve presented a principal biphasic profi le and in some cases minor 110 components were detected, which is more evident Arc with samples Frs, Aix and Ens. Th e complete com- Buz Frs 100 Cdl bustion of the resins took place near 600°C. Sal Fou Aix

90 DIFFERENTIAL THERMOGRAVIMETRIC (DTG) Age (Ma) Age EnsPlc Bel ANALYSIS OF THE FOSSIL RESINS 80 y = 0.6173x - 158.01 Figure 1 (thin lines) also shows the DTG analysis pat- r 2 = 0.5962 r = 0.7721 terns of combustion of the fossil resins investigated. 70 p = 0.0089 All samples were characterized by a main exothermal event, evidenced by the principal peak detected from 60 the beginning of combustion, and corresponding to 380 390 400 410 420 430 Main DTG peak (°C) the maximal rate of weight loss. Th e temperature of the main peak, reported also in Table 1, varied from FIG. 2. — Correlation between the main peak of Differential Ther- 394°C for sample Ens, to 420°C for samples Frs and mogravimetric (DTG) thermal event and the geological age of samples investigated. The regression is statistically signifi cant Aix. As shown by the scatter-plot and linear regres- (p = 0.0089). Vertical lines indicate the range of age of samples. sion of Figure 2, the increasing value of the main Abbreviations refer to samples as indicated in Table 1. exothermal peak correlated well to the increase of the age of the specimen, with a signifi cant correla- data. Th is mathematical procedure transforms a tion coeffi cient r of 0.7721 (p = 0.0089). number of possibly correlated variables into a smaller Figure 3 compares the data obtained from the number of uncorrelated variables; these so-called French amber specimens with those previously de- principal components are linear combinations of termined with samples of diff erent age and origin the original variables; once obtained, the principal (Ragazzi et al. 2003). Figure 3A presents the overall components are graphically plotted in order to data (n = 23) with the regression line constructed observe any groupings in the data set. PCA was to link the two parameters (age and main DTG determined with AMADA software (Xia & Xie peak), together with the calculated 95% confi dence 2001) on the matrix of DTG peak data, employ- limits. Th e correlation coeffi cient r was 0.6728 ing covariance matrix and standardized principal (p = 0.0004), suggesting a signifi cant association component scores. Th e fi rst three components were between the two parameters. used for the classifi cation of data. A 3D-scatter plot Being a theoretical construct on the experimental was obtained by means of KyPlot. data, and not a search of prediction of age through the determination of the main DTG peak, the in- ABBREVIATIONS verse plot was also determined by swapping the x See Table 1 for abbreviations of names of the am- and y axes. A new regression was calculated and the ber localities. theoretical values obtained this way were used to draw a new line. Th e average regression line (mean between the two previous lines) was also determined RESULTS and indicated in the plot of Figure 3B. Th e three THERMOGRAVIMETRIC (TG) PATTERNS lines suggest a possible range for the estimation of OF THE FOSSIL RESINS the link between the two variables of interest. Th ermogravimetric patterns of the French ambers are Th e main DTG peaks of ambers from France ap- presented in Figure 1 as thick lines. All the considered peared to fi t within the median range of the regres- samples were characterized by a TG combustion sion line and were quite well localized, according profi le which started after 200°C, corresponding to their age characteristic (see ellipse superimposed to the temperature at which a signifi cant weight on Figure 3A).

170 GEODIVERSITAS • 2009 • 31 (1) Th ermal analysis of Cretaceous French amber

As shown by the DTG profi le (Fig. 1), besides A the main exothermal event, additional peaks were 300 y = 2.169x - 808.72 2 found both before and after the main peak. Th e r r = = 0.4527 p0.4527r = 0.6728r = 0.6728 p = 0.0004 presence of these minor peaks is also reported in l m Table 1. One or two additional peaks following the main thermal event were generally observed for all 200 samples. Th e occurrence of early peaks, before the main exothermal event, was detected with samples Arc Buz Frs Frs Age (Ma) Age Arc Buz Aix Frs, Aix, Buz, Fou and Sal. Cdl Sal Aix 100 Ens Plc Th e whole pattern of peaks registered in DTG Ens Fou kk Bel Fou profi le was considered as a basis for possible fi nger- Bel j hh i f printing model of the French fossil resins. Th e dde i f e g amplitude of all the DTG peaks was measured and a b cc g 0 the values inserted in a matrix of data, which was 350 360 370 380 390 400 410 420 430 440 450 Main DTG peak (°C) furthermore used for Cluster Analysis and Principal Main DTG peak, °C Component Analysis. Th e results of CA shown in B Figure 4 as a dendrogram suggest the presence of 300 several clusters, the most diff erentiated being those including samples Arc and Cdl, Frs and Aix, and l m Buz and Sal, respectively. Another cluster, although with less diff erentiated elements, was that comprising 200 the samples Plc, and then Bel, Ens and Fou. Th e three-dimensional plot of the principal com- Arc Buz ponents calculated with PCA (Fig. 5) confi rmed (Ma) Age Frs Cdl Sal Aix the results obtained with CA, demonstrating four 100 Plc Bel Ens Fou k main groupings of the fossil resins. j h d i f ELEMENTAL ANALYSIS e a b c g Table 2 presents the relative abundance of the 0 diff erent elements analyzed in the 10 samples of 350 360 370 380 390 400 410 420 430 440 450 Main DTG peak (°C) French amber. Th e carbon content varied in a range of 73 to 80%, and hydrogen was between FIG. 3. — A, Correlation between the main peak of Differential 9.5 and 11.5%. Nitrogen was absent or appeared Thermogravimetric (DTG) thermal event and the geological age of the samples investigated, together with data obtained from amber only in traces (0 to 0.008%). Sulphur was gener- of different origin and age (data are from Ragazzi et al. 2003). The ally detected in small amounts, with highest levels dashed lines indicate 95% confi dence limits of the regression. The ellipse includes all samples from France, labelled as indicated in of the element in samples Buz, Fou and Plc, Bel, Table 1; B, lines indicate the plot obtained with different methods: Ens (with a range of 0.8 to 2.4%). regression of main DTG peak between age (full line), regression Th e amount of oxygen and trace elements was of age between main DTG peak (small-dashed line), average plot from both the previous regressions (large-dashed line). Letters in calculated as diff erence from the total. Th is value small case refer to the following samples used as a comparison: ranged between 4.6% in sample Plc to 16.8% in a, Picea abies resin; b, Madagascar copal; c, Colombia copal; d, blue Dominican amber; e, yellow Dominican amber; f, Mexican sample Sal. amber; g, Simetite; h, Lessini amber; i, Baltic amber; j, Cedar A multivariate approach was also tentatively used Lake amber; k, New Jersey amber; l, red Italian Triassic amber; with the data obtained with elemental analysis of m, Yellow Italian Triassic amber. the French amber samples. Th e dendrogram ob- tained with hierarchical CA (Fig. 6) does not show other specimens were less clearly diff erentiated, as clear clustering of the samples, with the greater indicated by the low distance among clusters. Th e diff erence observed for samples Sal and Plc. Th e three-dimensional plot of the principal components

GEODIVERSITAS • 2009 • 31 (1) 171 Ragazzi E. et al.

diff erent localities in southern France. Th e DTG Sal profi les show typically one main exothermal peak Buz Plc in the range of 394-420°C, which appeared to be Bel consistent with the thermal behaviour of other Ens coeval fossil resins previously investigated (Ragazzi Fou et al. 2003). Aix Our previous experience (Ragazzi et al. 2003) Amber samples Amber Frs suggested that thermal analysis can provide a valu- Cdl able way for classifying fossil resins, at least under Arc the aspect of age, since a progression of DTG exo- 0 5 10 15 20 25 30 thermal peak well correlates to the geological age Distance of the specimen. FIG. 4. — Dendrogram obtained with cluster analysis conducted Th e amberization process is quite complex since it using all the peaks detected from Differential Thermogravimetric comprises many diff erent events occurring in resin (DTG) analysis. Cluster analysis was performed with Ward’s method and non-standardized Euclidean distance. Abbreviations refer to fossilization after the initial water and monoterpene samples as indicated in Table 1. hydrocarbon loss: polymerisation of non-volatile components accompanied by cross-linking and isomerization reactions, account for resin matura- tion and several variables may infl uence the entire Frs process (Langenheim 1969, 2003; Pike 1993; Aix Anderson & Crelling 1995). Th e presence of a 2 correlation between the age of amber and its ther- mal behaviour provides further information on 1 Sal the history of the fossil resin, especially when the Fou Buz chemical composition is similar and the diagenetic 0 Plc Ens Arc variables are predominant. Being Late Albian or PC3 Bel -1 Cdl Early Cenomanian, the amber samples from the 1 Charentes investigated here (samples no. 1-5, see -1 Table 1) are confi ned within a relatively narrow 0 0 range of age. Th erefore it could be assumed that a -1 PC2 PC1 1 similar profi le of thermal analysis would have been obtained. Th is was overall confi rmed, but some dif- 2 -2 ferences both in main DTG peak and secondary FIG. 5. — Principal Component Analysis (PCA) carried out using peaks suggest that samples may have undergone a all the Differential Thermogravimetric (DTG) peaks detected with the 10 amber samples from France. PCA was obtained with co- diff erent maturation history. Th ermal analysis re- variance standardized matrix. Abbreviations refer to samples as vealed that samples from Archingeay and Cadeuil indicated in Table 1. (Arc and Cdl) are similar to each other, as are samples from Fouras and Île d’Aix (Frs and Aix), but the calculated with PCA (Fig. 7) confi rmed the results DTG profi les of both groups present peculiarities obtained with CA, demonstrating poor diff erentia- which allow a distinction. Th e lower main DTG tion among the specimens. peak suggests that samples from Archingeay and Cadeuil might be relatively younger than samples from Fouras and Île d’Aix, which is in contrast to DISCUSSION the stratigraphical data. However, samples from Fouras and Île d’Aix showed an early peak near Th e present data show the typical pattern of thermal 340°C, which is completely absent in samples from analysis of French Cretaceous ambers collected in Archingeay and Cadeuil. Th is early peak may be

172 GEODIVERSITAS • 2009 • 31 (1) Th ermal analysis of Cretaceous French amber

linked to the presence of more volatile components, Sal as in younger resins, a fact which is apparently Plc in contrast with fi nding a following higher main Aix exothermal peak. A possible explanation of these Frs discrepancies may involve the presence of fossil Buz resins originated from diff erent plants, containing Cdl components with diff erent characteristics of ther- Fou Amber samples Amber Ens mal degradation. According to Perrichot (2005), Bel ambers from Charentes would originate from the Arc conifer family Araucariaceae as the main botanical 0246810141612 source. However, since FTIR analysis on various Distance specimens also produced diff erent spectra, the possibility of more than one source is likely, and FIG. 6. — Dendrogram obtained with cluster analysis conducted using elemental analysis. Cluster analysis was performed with the plant remains associated in the diff erent amber Ward’s method and non-standardized Euclidean distance. Ab- sites (Gomez et al. 2004; Néraudeau et al. 2005) breviations refer to samples as indicated in Table 1. also suggest a Cheirolepidiaceae, Podocarpaceae or Taxodiaceae as possible resin producers. Th e older the resin is, the lower are the additional peaks. Th is is a general pattern emerging from the comparative analysis of previously published data 2 (Ragazzi et al. 2003). Th is fact has been particu- ▲Fou Plc Ens ▲ larly evident with Italian Triassic amber, the oldest 1 ▲ Aix Bel▲▲▲ among the studied resins, which presented only ▲ Frs one relevant exothermal event (with the maximum 0 ▲ Arc Buz ▲ Cdl peak at 437-443°C), and a narrow secondary peak PC3 -1 ▲Sal after 500°C. Th e French ambers from Charentes, -2 -1 and particularly samples Arc and Cdl, presented 1 0 secondary peak(s) in the range 450-600°C, lower 0 1 than those of the other samples (Fig. 1). Th is fact -1 PC1 2 suggests that all other studied samples are younger -2 PC2 than samples Arc and Cdl, as supported by strati- 3 graphical data (Table 1). Th is may indicate that these ambers were not redeposited from older sediments FIG. 7. — Principal Component Analysis (PCA) conducted on el- emental analysis of the 10 amber samples from France. PCA was and that the amber formation and sedimentation obtained with covariance standardized matrix. Abbreviations refer were (at least almost) contemporaneous. to samples as indicated in Table 1. Th e data obtained from all French localities fi t well within the overall information provided by thermal analysis conducted on samples from all Another result emerged from the multivariate over the world and diff erent age. Th e correlation techniques of data analysis. Th e use of CA in the age vs main thermal peak was confi rmed, also when present investigation permitted to confi rm math- the two variables are swapped, suggesting that the ematically what has been indicated above after a rate of weight loss displayed by the fossil resin can general inspection of thermal analysis spectra. A be satisfactorily used as a marker of the age of the quite similar DTG pattern was observed between resin. Th e correlation showed outliers, and this may samples from Archingeay and Cadeuil (Arc and depend from the fact that, as discussed earlier, age Cdl), as well as between samples from Fouras and is not the only determinant of the thermal behav- Île d’Aix (Frs and Aix), or between samples from iour of the resin. La Buzinie and Salignac (Buz and Sal), and among

GEODIVERSITAS • 2009 • 31 (1) 173 Ragazzi E. et al.

the group of samples from Piolenc, Belcodène, Acknowledgements Ensuès-la-Redonne and Fourtou (Plc, Bel, Ens and We thank all colleagues and persons who helped Fou). Also PCA led to the same result confi rming us in fi nding some of the amber deposits studied the possibility to obtain a fi ngerprint of the fossil here: Alain Couillard, Christian Devalque, Luc resins by using a multivariate approach. Ebbo, Ginette Lambert, Jean Le Loeuff , André As a further consideration, samples Plc, Bel and Nel, Gaël de Ploëg, Daniel Roggero. We thank Ens, coming from Vaucluse and Bouches-du-Rhône Kerstin Schmidt for the help in the preparation and appearing located within a cluster, suggest a of the manuscript. Th is work is a contribution to similar palaeobotanical origin and/or taphonomic the project AMBRACE no. BLAN07-1-184190 conditions. Th ese samples are Santonian in age, and from the Agence nationale de la Recherche (ANR). according to Gomez et al. (2003) the region was Support was also provided by the Alexander von characterized by a rich fl ora with large abundance Humboldt Foundation (to V.P.), by the German in angiosperms in ancient times, and a diff erent Research Foundation (to A.R.S.), by C.N.R. Insti- palaeobotanical origin could be the reason for a tute of Geosciences and Earth Resources, Padova, diff erent thermal profi le of the resin, in comparison Italy (to G.R.) and by the M.I.U.R. Italy, grant no. to the other samples. MM04233815/003-De Zanche (to G.R.). As shown also by the mean values of Table 2, the fossil resins presented an overall composition which is consistent with that of other samples from REFERENCES other sites of the world and diff erent age (Ragaz- zi et al. 2003). Carbon revealed to be the main ANDERSON K. B. & CRELLING J. C. (eds) 1995. — constituent, with a range between 73 and 80% of Amber, Resinite, and Fossil Resins. ACS Symposium the total, without any relation with the age of the Series 617, American Chemical Society, Washington samples. It is noteworthy that the sulphur content D.C., 297 p. ANDERSON K. J. & WINANS R. E. 1991. — Nature and was higher in samples from Piolenc, Belcodène and fate of natural resins in the geosphere. 1. Evaluation of Ensuès-la-Redonne (Plc, Bel, and Ens) and from pyrolysis-gas chromatography/mass spectrometry for La Buzinie and Fourtou (Buz and Fou), in partial the analysis of natural resins and resinites. Analytical agreement with the clustering found on the basis Chemistry 63: 2901-2908. of DTG behaviour. A higher sulphur content has ANDERSON K. B., WINANS R. E. & BOTTO R. E. 1992. — Th e nature and fate of natural resins in the geosphere. been found also in Triassic amber from Dolomites II. Identifi cation, classifi cation and nomenclature of (Ragazzi et al. 2003), and this fact may depend resinites. Organic Geochemistry 18: 829-841. on the diff usion of the element present in the BECK C. W. 1986. — Spectroscopic investigations of embedding sediment to the resin. An attempt to amber. Applied Spectroscopy Reviews 22: 57-110. fi nd a clustering of the samples on the basis of the BROUGHTON P. L. 1974. — Conceptual frameworks for geographic-botanical affi nities of fossil resins. Cana- elemental composition did not achieve appreciable dian Journal of Earth Sciences 11: 583-594. results, using both CA and PCA. Most probably CARLSEN L., FELDTHUS A., KLARSKOV T. & S HEDRINSKY the result depends on the fact that the elemental A. 1997. — Geographical classifi cation of amber based composition was not diff erent enough among the on pyrolysis and infra-red spectroscopy data. Journal studied samples of French amber. of Analytical and Applied Pyrolysis 43: 71-81. DEJAX J. & MASURE E. 2005. — Analyse palynologique In conclusion, thermal analysis is confi rmed to de l’argile lignitifère à ambre de l’Albien terminal be a useful source of information also for fossil d’Archingeay (Charente-Maritime, France). Comptes resins from France, that may help to identify age, Rendus Palevol 4 (1-2): 167-177. diagenetic processes and palaeobotanical origin of FEIST M., LAMPRECHT I. & MÜLLER F. 2007. — Th ermal the amber. Th e use of multivariate data analysis, investigations of amber and copal. Th ermochimica Acta 458: 162-170. based on the data provided by DTG spectra, makes GOMEZ B., BARALE G., SAAD D. & PERRICHOT V. 2003. — it possible to build a reliable fi ngerprint for each Santonian Angiosperm-dominated leaf-assemblage fossil resin. from Piolenc. Comptes Rendus Palevol 2: 197-204.

174 GEODIVERSITAS • 2009 • 31 (1) Th ermal analysis of Cretaceous French amber

GOMEZ B., DAVIERO-GOMEZ V., PERRICHOT V., THÉVE- new amber deposit from the Cretaceous (uppermost NARD F., COIFFARD C., PHILIPPE M. & NÉRAUDEAU Albian-Lowermost Cenomanian) of SW France. D. 2004. — Assemblages fl oristiques de l’Albien- Creta ceous Research 29: 925-929. Cénomanien de Charente-Maritime (SO France). NÉRAUDEAU D., VULLO R., GIRARD V., GOMEZ B., Annales de Paléontologie 90: 147-159. PERRICHOT V. & VIDET B. 2009. — Amber, plants GRADSTEIN F. M., OGG J. G., SMITH A. G. (eds) 2004. — and vertebrates from the paralic Lower Cenomanian A Geologic Time Scale 2004. Cambridge University deposits of Aix Island (Charente-Maritime, France). Press, Cambridge, 585 p. Geodiversitas 31 (1): 13-28. GRIMALT T. O., SIMONEIT B. R. T., HATCHER P. G. & PERRICHOT V. 2005. — Environnements paraliques NISSENBAUM A. 1988. — Th e molecular composition à ambre et à végétaux du Crétacé Nord-Aquitain of ambers. Organic Geochemistry 13: 677-690. (Charentes, Sud-Ouest de la France). Mémoires de LAMBERT J. B. & FRYE J. S. 1982. — Carbon functional- Géosciences Rennes 118: 1-310. ity in amber. Science 217: 55-57. PERRICHOT V., NEL A. & NÉRAUDEAU D. 2004. — Two LANGENHEIM J. H. 1969. — Amber: a botanical inquiry. new wedge-shaped beetles in Albo-Cenomanian ambers Science 163: 1157-1169. of France (Coleoptera: Ripiphoridae: Ripiphorinae). LANGENHEIM J. H. 2003. — Plant Resins: Chemistry, European Journal of Entomology 101 (4): 583-589. Evolution, Ecology, and Ethnobotany. Timber Press, PERRICHOT V., NEL A. & NÉRAUDEAU D. 2007a. — Portland, 586 p. Schizopterid bugs (Insecta: Heteroptera) in mid- MARTÍNEZ-RICHA A., VERA-GRAZIANO R., RIVERA A. & Cretaceous ambers from France and Myanmar (Bur- JOSEPH-NATHAN P. 2000. — A solid-state 13C NMR ma). Palaeontology 50 (6): 1367-1374. analysis of ambers. Polymer 41: 743-750. PERRICHOT V., NÉRAUDEAU D., NEL A. & DE PLOËG NÉRAUDEAU D., PERRICHOT V., DEJAX J., MASURE E., G. 2007b. — A reassessment of the Cretaceous NEL A., PHILIPPE M., MOREAU P., GUILLOCHEAU F. & amber deposits from France and their palaeonto- GUYOT T. 2002. — Un nouveau gisement à ambre logical signifi cance. African Invertebrates 48 (1): insectifère et à végétaux (Albien terminal probable) : 213-227. Archingeay (Charente-Maritime, France). Geobios PEYROT D., JOLLY D. & BARRON E. 2005. — Ap- 35: 233-240. port de données palynologiques à la reconstruction NÉRAUDEAU D., ALLAIN R., PERRICHOT V., VIDET B., DE paléoenvironnementale de l’Albo-Cénomanien des LAPPARENT DE BROIN F., GUILLOCHEAU F., PHILIPPE Charentes (Sud-Ouest de la France). Comptes Rendus M. & RAGE J.-C. 2003. — A new Early-Cenomanian Palevol 4 (1-2): 151-165. paralic deposit with fossil wood, amber with insects PIKE E. M. 1993. — Amber taphonomy and collecting and Iguanodontidae (Dinosauria, Ornithopoda) at bias. Palaios 8: 411-419. Fouras (Charente-Maritime, southwestern France). RAGAZZI E., ROGHI G., GIARETTA A. & GIANOLLA P. Comptes Rendus Palevol 2: 221-230. 2003. — Classifi cation of amber based on thermal NÉRAUDEAU D., VULLO R., GOMEZ B., PERRICHOT analysis. Th ermochimica Acta 404: 43-54. V. & VIDET B. 2005. — Stratigraphie et paléonto- RODGERS K. A. & CURRIE S. 1999. — A thermal analyti- logie (plantes, vertébrés) de la série paralique Albien cal study of some modern and fossils resins from New terminal – Cénomanien basal de Tonnay-Charente Zealand. Th ermochimica Acta 326: 143-149. (Charente-Maritime, France). Comptes Rendus Palevol XIA X. & XIE Z. 2001. — AMADA: analysis of micro- 4 (1-2): 79-93. array data. Bioinformatics 17: 569-570. NÉRAUDEAU D., PERRICHOT V., COLIN J.-P., GIRARD V., YOSHIOKA K. 2002. — KyPlot – A user-oriented tool GOMEZ B., GUILLOCHEAU F., MASURE E., PEYROT for statistical data analysis and visualization. Compu- D., TOSTAIN F., VIDET B. & VULLO R. 2008. — A tational Statistics 17: 425-437.

Submitted on 4 October 2007; accepted on 23 September 2008.

GEODIVERSITAS • 2009 • 31 (1) 175